I so wanna know more about this, I'll be spending a chunk of the afternoon reading on this story.
The future used to be so cool. Having access to Wikipedia and live reporting always on hand was going to make everyone so informed, easier communications were going to make democracy and grassroots organizing flourish... We were all going to have our own server on our pockets...
We were naive, but I sorely miss the tech-optimism that to some extent carried me through uni.
A further example of how stunningly naive this was: they wanted (external NMEA) GPS connected to it (and Bluetooth beacons when indoors, which had to be created for the project) so people could publish their real time location.
They say it's a waste that you pay for the tokens and then the inference provider pays for the electricity. Isn't that how everything works? I pay cucumbers and the farmers have to pay for the water and the fertilizer...
I hope that reasoning is an after-the-fact justification by the LLM that wrote this.
It's a ver interesting idea and I wouldn't mind trying it out, but with a smaller model. At 0.5t/s and reading many gigabytes from the SDD every second... I wonder if this wouldn't be extremely practical if targeting a 500gib or 250gib model, something that is still outside most consumers' laptop.
It is worth thinking about, when you're buying something, how much of what you're paying is the supplier's margin. And whether they have efficiency advantages over you doing it yourself. But for large LLMs it does seem like there's a pretty big efficiency advantage to the rack-scale hardware in datacenters compared to hacks like this.
here at 0.5 tok/s, it's like getting 1/100 of a tomato for your lunch per hour, and you can't just let it grow, you have to look at it growing just as you're about to eat, and it doesn't accumulate between meals
That'd be a huge performance boost too, the top end Snapdragons probably easily beat the Wildcat Lake chips but I believe they're still catching up on Linux support. Wildcat Lake would also let you put a Thunderbolt device on it which can also level-up its GPU capabilities.
It still happens, problems that are solved still happen when people don't take care to apply the solution. Diseases that were solved problems happen again when people stop taking the vaccines.
You can avoid SQL injection by just coding the same features with a bit of care. You loose nothing. Mistakes can always happen, but it's not even tricky to prevent SQL injection.
Right now the only way to avoid Prompt injection is to not let your agents see user input at all. A very wide range of features that we'd like to implement are unsafe and there isn't a way to prevent this reliably.
I guess we'll need to get used to control the agent's permissions very tightly, and taylor them per-conversation. The agent I speak to for customer support must only have access to my data, and not because of instructions in the system prompt, these will need to be hard limits.
Jumped to GrapheneOS a few months ago. Works great. The keyboard was bad but you should install FUTO, as some other comments recommend.
My only issue with it has been a few apps not working correctly, and not the ones I expected. I did my research before hand and knew that my banking apps would work, thinking those would be the main challenge.
Turns out the bike-sharing system in my city, Madrid, won't work. I ended up installing Google Play services (that run sandboxed in Graphene, but still wanted to avoid), and it works sometimes, but mostly doesn't. I use these bikes a few times a week, so this is a major hassle, and I end up carrying my ancient iPhone with me sometimes just for this.
This and Trade Republic have been my only two problems. Happy otherwise, but do your research before switching, and don't assume only the apps you expect to be problematic will be.
Why don't you connect with the makers of Madrid and see what they can do about it? That is sometimes the best way to fix these types of incompatibilities.
Sadly, this will always be a game of cat and mouse. I wonder if it was ever avoidable for the internet to become a hundred of such games at once, between advertisers and ad blockers, trackers and browsers, hackers and developers, search engines and SEO experts...
If your concept took off, someone will make a program that inputs the LLMs text like a human would. It might still take some time (if it isn't possible to hack the timing records client-side), you'd probably have to get busy detecting multiple connections from the same IP, and then they'd work around it...
Qwen 3.7 Max:
> During my local testing before the full eval harness it was the only non-GPT model that was able to complete the task, was not able to reproduce in the longer runs.
Doesn't that sound like may be the harness was the problem?
I was using the same harness for each run, the difference is from when I was running the harness locally on my machine before I pushed up the full runs.
It could just alias sudo on your ~/.bashrc. No need to replace the actual file on /usr/bin/sudo or wherever you have it. I would only need to be able to run arbitrary code as you.
I so wanted to love the Steam Deck, but it's a device with a 7 inch screen that occupies a massive volume on your bag. Unless you know you're going to play a fair ammount, it's not worth carrying around.
It's a fantastic console, but a mediocre general purpose computer.
I have one, and I love it! I don't use it as much as I thought I would, but it brings joy everytime. If you have a need for an extremely portable, not very powerful device on your life, this might be it.
I agree with the complaint about the trackpad, but the keyboard has been just fine for me. Just a bit small, of course. I also find the screen perfectly acceptable for what I use this thing for: youtube, taking notes, writing emails, small bouts of coding and ssh'ing into servers.
My main complaint is related to battery management. May be it's becaused I'm used to Macbooks, but it drives me nuts to go pick the Minibook up and find that it has no power, because I haven't used it in a couple of days and I put it to sleep. I haven't measured, but the power use on sleep is noticeable, and I suspect the leakage while hibernating might be significant too.
I don't really like the laptop form factor. Laptops are the perfect solution for only one use case: using them on your lap. On a table, I'd rather have the computer be just a tablet, to add a bluetooth keyboard and mouse. At my desk, with bigger screens, I'd like the computer to disappear into a small puck or box, like a Mac Mini. With the Minibook, being so small, the form factor makes sense again. It's so portable, so easy to take with me to a coffee shop or on a trip, it's worth it.
A tablet with a keyboard might be a more practical solution, although generally more expensive, but I appreciate that my Minibook runs Linux so well, so I don't have to even think about Apple or Google telling me how to use my computer.
The future used to be so cool. Having access to Wikipedia and live reporting always on hand was going to make everyone so informed, easier communications were going to make democracy and grassroots organizing flourish... We were all going to have our own server on our pockets...
We were naive, but I sorely miss the tech-optimism that to some extent carried me through uni.